A non-procedural approach might be easier: collect loops of baseline murmurs / elevated cheering / peak ecstatic crowd and vary the volume of each based on what is happening on screen. If you have variants of each "energy level" you can also smoothly fade between them, queuing the next one randomly as the current version runs out. This way the wrap-around point should not be that noticeable.
This is sometimes known as "walla". This tutorial is specific to Sound Particle (audio app), but the basics can be implemented easily from scratch. https://youtu.be/nigFrC2mORk?si=Cktdb3aTrIs5vUbB
For a game I was making years ago, I tried to make a “crowd noise” generator. As in for a football/soccer match.
But it was not my skillset and I failed miserably.
Does anyone know of something that might get me started?